pci_enable_device for pciaccess?

[Dave Airlie]

On Tue, May 20, 2008 at 1:26 PM, Ian Romanick <idr at us.ibm.com> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Dave Airlie wrote:
> | Okay I've been playing with getting posting working again in master,
> |
> | I can make it work on Linux if I use the OS to enable the device in
> | /sys before trying to POST it.
> |
> | However I really think this should be pciaccess functionality, so I'd
> | like to add a pci_enable_device function to libpciaccess and call it
> | from the server before trying to POST secondary cards.
> |
> | Long term the posting should be done before X starts but I want to fix
> | the regression for Fedora.
> |
> | the other option is to haxor the int10 module to do this on Linux, but
> | I'd rather expand pciaccess.
>
> Adding this functionality to libpciaccess is the right thing to do.
> When enabling a device, to the buses leading to the device also have to
> be enabled?  I was thinking about adding some functionality to iterate
> from the root to a particular device.  This would facilitate enabling
> the buses leading to the device and enabling VGA to a device.
>

I'm mainly assuming the bus is enabled to the device, which seems to
be the case in most cases on Linux or else the device wouldn't be
visible.

VGA bits I'm staying away from until we the VGA arbiter in some useful
form to at least prove an interface against.

All this function does is enable the device BARs, the Linux kernel
already support this via /sys.

Dave.